A semi-supervised Intrusion Detection System using active learning SVM and fuzzy c-means clustering

one of the most crucial tool for network defense is an Intrusion Detection Systems (IDS). Several supervised machine learning algorithms exist that deal with high volumes of training data. However, there is a need to design optimized, low cost semi-supervised IDS model that can perform well with little labelled data. The challenge is also to design adaptive learning algorithms that can update themselves with minimum computational overhead in case new training data is to be included. This work demonstrates a hybrid semi-supervised machine learning technique that uses Active learning Support Vector Machine (ASVM) and Fuzzy C-Means (FCM) clustering in the design of an efficient IDS. This algorithm is tested on NSL KDD bench mark IDS data set and found to be promising.